Registry-based randomised phase II study of Enzalutamide versus Abiraterone: assessing cognitive function in eLderly patients with metastatic castration-resistant Prostate cancer (REAL-Pro)

Abstract

Background

Androgen receptor pathway inhibitors (ARPIs) are commonly used in older patients with metastatic castration-resistant prostate cancer (mCRPC). Treatment selection is informed by considering adverse event (AE) profiles, given that there are no previous randomised phase III comparison trials.

Materials and methods

This prospective registry-based randomised trial utilised the electronic Prostate Cancer Australian Database to collect clinical data. Patients aged ≥75 years who were suitable to receive abiraterone or enzalutamide for mCRPC were randomised to receive either ARPI treatment, stratified by prior docetaxel use. Telephone assessments were conducted at baseline and 12 weeks, including the Blessed Orientation-Memory-Concentration (BOMC) tool, Geriatric Depression Scale (GDS), and Falls Risk Questionnaire (FRQ). Descriptive statistics and mixed-effects linear regression were used to compare groups, including changes in assessment scores from baseline. Cox proportional hazards modelling was used to analyse effects of variables on time to treatment failure (TTF) and overall survival (OS).

Results

We enrolled 76 men between June 2019 and September 2023, but closed due to slow accrual in December 2023. Nineteen (25%) had prior docetaxel. The median age was 81 years (interquartile range 77-85 years). The mean score change at 12 weeks trended higher in those receiving enzalutamide for each of BOMC (+0.64 versus −0.51, P = 0.46), GDS (+0.70 versus +0.06, P = 0.06), and FRQ (+1.4 versus −0.12, P = 0.06). A higher baseline GDS score was associated with shorter TTF [hazard ratio (HR) 1.3, 95% confidence interval (CI) 1.05-1.6, P = 0.014], as was higher baseline FRQ score (HR 1.12, 95% CI 1.09-1.22, P = 0.012). There were no differences in TTF, OS, prostate-specific antigen response, or AE rates between treatment groups.

Conclusions

Our results highlight the importance of individualised assessment of cognition, depression, and falls risk in older patients commencing ARPI treatment.

Highlights

• •This was a pragmatic randomised controlled trial embedded in a real-world clinical registry.
• •The study compared patient-reported outcomes in older patients with prostate cancer.
• •Patients receiving enzalutamide had a trend towards worsening scores after 12 weeks.
• •Baseline falls risk and depression scores correlated with long-term clinical outcomes.

Introduction

Prostate cancer is the most common non-cutaneous malignancy in men worldwide,¹ with its incidence increasing with age.² The median age of prostate cancer diagnosis in Australia is 69 years.² Those who are ≥75 years of age are typically underrepresented in conventional clinical trials, with a Food and Drug Administration analysis finding that men >75 years of age represented only 12% of trial participants, despite comprising 30% of new cancer diagnoses.³ There are therefore limited trial data to inform the expected benefits and risks of adverse events (AEs) for many patients treated in the real-world setting.

Within our Australian real-world prostate cancer registry, 43% of men diagnosed with metastatic castration-resistant prostate cancer (mCRPC) were aged ≥75 years.⁴ These older patients commonly receive androgen receptor pathway inhibitors (ARPIs) for mCRPC, with treatment selection often influenced by baseline comorbidities. While several retrospective real-world studies have suggested higher prostate-specific antigen (PSA) response rates, and in some instances improved survival with enzalutamide compared with abiraterone,5, 6, 7 there is no randomised trial comparing these agents in the older population.

Patient-reported outcomes and quality of life are particularly important in the older population due to the elevated risk of adverse outcomes.^8,9 Previous studies have demonstrated a decline in objective and subjective measures of cognitive function as early as 3 months following the commencement of enzalutamide treatment.¹⁰ Inferior health-related quality-of-life measures have also been demonstrated in patients >75 years of age who receive enzalutamide compared with abiraterone.¹¹ Although comprehensive geriatric assessments (CGAs) have demonstrated the potential to predict toxicity and prevent hospitalisations in older adults with cancer, they are resource intensive and have not been widely integrated into routine oncology practice.^12,13 As such, the optimal assessments to guide treatment decision making in the older population remain unclear.

The registry-based randomised controlled trial (RRCT) is an emerging methodology that allows greater representation of a real-world population, increasing external validity and generalisability with pragmatic designs and broad eligibility.¹⁴ This is particularly attractive for questions related to older patients, who are underrepresented in standard trials despite receiving study agents in routine practice. While clinical registries are commonly used in oncology to capture real-world data, analyse treatment patterns, and compare outcomes, the absence of randomisation introduces potential selection bias and confounding factors that impact outcomes. RRCTs, by incorporating randomisation, allow quality prospective trial designs in a real-world population. These trials are embedded within an existing registry and therefore have significantly lower cost and resource requirements.

The Registry-based Randomised Study of Enzalutamide versus Abiraterone Assessing Cognitive Function in Elderly patients with Metastatic Castration-Resistant Prostate Cancer (REAL-Pro) is an RRCT utilising the existing electronic Prostate Cancer Australian Database (ePAD) for prospective data collection.¹⁵ This pragmatic study aimed to address an important clinical question in a real-world population of older patients with mCRPC. The Australian New Zealand Clinical Trials Registry registration number for this study is ACTRN1 2620000463976.

Materials and methods

Patients aged ≥75 years with mCRPC who were deemed suitable to receive full-dose abiraterone or enzalutamide by their treating clinician, both widely available via the Australian Pharmaceutical Benefit Scheme, were randomised to either treatment. At the time of enrolment, apalutamide or darolutamide was not reimbursed for mCRPC in Australia and no ARPIs were reimbursed for hormone-sensitive prostate cancer (HSPC). Patients were stratified by prior use of docetaxel chemotherapy for HSPC or mCRPC. Patients were excluded if they had prior systemic therapy for mCRPC other than docetaxel or if they were unable to complete telephone assessments. Following informed consent, randomisation was carried out in real time by a central study coordinator upon contact from the treating clinician and confirmation of eligibility. A phone interview was conducted at baseline (within 7 days of commencing treatment) and after 12 weeks (±7 days) by the central study coordinator using validated tools: Blessed Orientation-Memory-Concentration (BOMC) tool, Geriatric Depression Scale (GDS), and Falls Risk Questionnaire (FRQ).

The BOMC is a six-item scoring tool for assessing cognitive function, producing a continuous score.¹⁶ It has been validated for administration over the telephone and its use is endorsed in older patient populations within several international guidelines.^17,18 It takes 2-3 min to conduct and utility was demonstrated in a cancer-specific geriatric assessment feasibility study.¹⁹ The FRQ 12-item score is highly predictive of falls risk based on clinical examination, validated in an oncology setting, and correlates with quality-of-life measures and the more resource-intensive ‘timed up-and-go’ test.^20,21 The GDS, a 15-item self-reporting instrument, has high correlation with depressive symptoms and is a recommended screening tool for older adults with cancer in international guidelines.^22,23

The primary objective was to compare cognitive decline based on changes in BOMC scores between treatment groups. Secondary objectives were to compare depression rates using GDS, the incidence of falls, serious AEs, PSA50 response rates, time to treatment failure (TTF), and overall survival (OS). PSA50 response rate was defined as the proportion of patients achieving ≥50% reduction in PSA from baseline during therapy. TTF was defined as the time from treatment commencement to cessation for progressive disease (either radiological, clinical, or biochemical, based on the discretion of the treating clinician). Where treatments were stopped for reasons other than progression, such as toxicity, TTF was defined as the time from treatment commencement to the commencement of subsequent systemic therapy, following disease progression. Participants who had no disease progression were censored at the time of last follow-up. OS was defined as the time from randomisation until the date of death.

Patients who were unable to complete phone assessments at the time of initial contact were given the opportunity to nominate a suitable time on two subsequent occasions. All patients completed a baseline assessment. Patients who did not complete a 12-week assessment were still included in the study and follow-up clinical data were collected. In case distress could be triggered by the assessments, 24-h contact details for mental health national support organisation Beyond Blue and crisis-support and suicide prevention organisation Lifeline Australia were provided in the patient consent form and during telephone interviews. If patients recorded clinically significant scores potentially warranting timely clinical intervention, based on validated thresholds, the treating clinician was informed within 24 h of assessment.

The frequency of clinical reviews and blood and imaging investigations were carried out at the discretion of the treating site clinician, as per standard practice. A PSA within 7 days of enrolment was mandatory. Patients commenced treatment at the full recommended dose: abiraterone 1000 mg daily with concurrent corticosteroid or enzalutamide 160 mg daily. Subsequent dose alterations were at the discretion of the treating clinician. Management decisions, including changing systemic therapy, were also at the discretion of the treating clinician. Data obtained from the telephone assessments were recorded in a secure REDCap database. All other clinical data were collected using established methods of the ePAD.¹⁵ AEs collected in the database are limited to those that are clinically notable, defined as an event that results in either hospitalisation, dose reduction, or delay.

The ePAD registry collects clinicopathological information from consecutive patients at multiple sites within Australia and data are entered through review of all available medical records by trained data abstractors into a password-protected web-based interface.¹⁵ Sites that were not involved in ePAD at the time of study commencement were encouraged to join the ePAD registry to enable data entry or were offered the option of using paper-based data collection forms. Paper forms were used at two sites, with data subsequently merged with ePAD data.

The planned total recruitment was 200 patients, predicted to be achieved over 36 months based on previous ePAD enrolment rates of patients who appeared to meet the eligibility criteria. A previous study had demonstrated objective reduction in cognitive function in 19% of patients on enzalutamide and 5% of patients on abiraterone after 12 weeks of therapy.²⁴ A Fisher’s exact test with a 0.05 two-sided significance level was predicted to have 80% power to detect the difference between a proportion of 0.05 (abiraterone) and 0.19 (enzalutamide) with a sample size of 90 in each group. Adjusted for 10% loss to follow-up, 100 patients were planned to be randomised in each arm.

Recruitment commenced in June 2019. The study closed to recruitment early, in December 2023, due to slow accrual and the changing standards of care. ARPIs became widely available for use for HSPC in April 2023, meaning fewer patients developing castration resistance were ARPI-naive. Subsequent analyses from ePAD also suggested that the real-world incidence of comorbidities and potential drug interactions were significantly higher than anticipated,²⁵ reducing the number of eligible patients. Given the significant reduction in sample size, further exploratory analyses utilised raw assessment scores, rather than comparing the number of patients with score changes. A worsening in BOMC, GDS, or FRQ score was defined as an increase of the total score by ≥1 since baseline. Differences between baseline and 12-week scores were calculated for each patient and means for each treatment group were compared for each of BOMC, GDS, and FRQ assessments.

Descriptive statistics were used to report patient and disease characteristics. Treatment groups were compared using Fisher’s exact tests, chi-square analyses, and t-tests. Time-to-event analyses were carried out using Kaplan–Meier methods. Random-intercepts mixed-effects linear regression was used for exploratory analysis to compare score changes between treatment groups and to analyse the use of prior docetaxel on subsequent scores at 12 weeks. Cox proportional hazards univariate and multivariate models were used to analyse the effects of scores and clinical variables on TTF and OS. Variables with a P value of <0.1 on univariate analysis were included in the multivariable model. The primary endpoint and clinical outcomes were analysed in the entire intention-to-treat population. TTF and OS were calculated from the 12-week mark when assessing 12-week scores in a separate Cox proportional hazards model, including only those who completed the 12-week assessments. Statistical analyses were carried out using Stata BE version 18.0 (2023; StataCorp LLC, College Station, TX).

Results

We enrolled 76 patients with mCRPC between 12 June 2019 and 11 September 2023. A total of 40 patients were randomised to abiraterone and 36 to enzalutamide. Baseline characteristics are displayed in Table 1. Median age was 81 years overall (interquartile range 77-85 years), 81 years in the abiraterone group, and 80 in the enzalutamide group (P = 0.33). Eighteen (24%) had received prior docetaxel, including 10 (25%) in the abiraterone group and 8 (22%) in the enzalutamide group. More patients receiving abiraterone had diabetes (25% versus 3%, P = 0.03) but other characteristics appeared well matched, including time to CRPC, the proportion with de novo metastases, and median PSA at CRPC diagnosis.

Table 1.

Baseline characteristics

Treatment	Abiraterone n = 40	Enzalutamide n = 36	P value
Median age, years	81	80	0.33
Hypertension, n (%)	24 (60)	23 (64)	0.31
Dyslipidaemia, n (%)	19 (48)	15 (42)	>0.99
Diabetes, n (%)	10 (25)	2 (6)	0.03
Ischaemic heart disease, n (%)	1 (2.5)	2 (6)	0.59
Median time to CRPC, months	42.7	43.4	0.96
De novo metastases, n (%)	21 (53)	17 (47)	0.63
Visceral metastases, n (%)	5 (12.5)	0 (0)	0.06
Median PSA, ng/ml	62	38	0.43
Prior docetaxel, n (%)	10 (25)	8 (22)	>0.99

CRPC, castration-resistant prostate cancer; PSA, prostate-specific antigen.

Fifty-six (74%) patients completed both telephone assessments (28 in each group). Of the patients who did not complete the second assessment (n = 20), six had stopped ARPI treatment before 12 weeks, including one who had died, six lost to follow-up, and eight had ‘other’ reasons documented, including patient preference.

There were no statistically significant differences in mean scores between treatment groups for any assessment tool at baseline or 12 weeks (Figure 1). After 12 weeks, mean FRQ score was numerically higher in those receiving enzalutamide (5.6 versus 4.0; difference −1.6, 95% CI −3.6 to +0.47, P = 0.12). A numerically greater proportion of patients on enzalutamide had worsening BOMC (39% versus 29%, P = 0.57) and GDS (46% versus 29%, P = 0.27) (Table 2). In the mixed-effects linear regression model, the mean change in individual scores over 12 weeks was numerically higher, indicating worsening, in those receiving enzalutamide for BOMC (+0.64 versus −0.51, P = 0.46), GDS (+0.70 versus +0.06, P = 0.06), and FRQ (+1.4 versus −0.12, P = 0.06) (Figure 2; Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2025.100677). The use of prior docetaxel was associated with neither baseline scores nor a decrease in 12-week scores (Supplementary Table S1, available at https://doi.org/10.1016/j.esmorw.2025.100677).

Figure 1.

Mean scores at baseline and 12 weeks. BOMC, Blessed Orientation-Memory-Concentration; GDS, Geriatric Depression Scale; FRQ, Falls Risk Questionnaire.

Table 2.

Proportion of patients with worsening assessment scores at week 12

	Abiraterone n = 28 n (%)	Enzalutamide n = 28 n (%)	P value
BOMC	8 (29)	11 (39)	0.57
GDS	8 (29)	13 (46)	0.27
FRQ	14 (50)	15 (54)	>0.99

BOMC, Blessed Orientation-Memory-Concentration; GDS, Geriatric Depression Scale; FRQ, Falls Risk Questionnaire.

Figure 2.

Mean change in scores from baseline. BOMC, Blessed Orientation-Memory-Concentration; GDS, Geriatric Depression Scale; FRQ, Falls Risk Questionnaire.

There were no differences in PSA50 response rates for those receiving abiraterone or enzalutamide, respectively (50% versus 39%, P = 0.75; Table 3). TTF was not statistically different between groups [hazard ratio (HR) 0.99, 95% confidence interval (CI) 0.50-1.97, P = 0.98], nor was OS (HR 1.27, 95% CI 0.61-2.64, P = 0.98) (Figure 3). The AE rate was also similar between groups (37.5% versus 31% for abiraterone versus enzalutamide, P = 0.79). Falls occurred in six (21%) patients on enzalutamide and three (11%) on abiraterone (P = 0.29). The 12-week FRQ score was significantly associated with falls [odds ratio (OR) 1.24, 95% CI 1.04-1.48, P = 0.02] but baseline FRQ scores were not (OR 1.14, 95% CI 0.96-1.35, P = 0.12). In total, 43 patients had stopped ARPI treatment at the time of last follow-up. Most patients (n = 34) ceased treatment due to disease progression, but nine patients ceased treatment due to toxicity, including six (21%) on abiraterone and three (11%) on enzalutamide. Five of these subsequently changed to the alternate ARPI.

Table 3.

Treatment outcomes

	Abiraterone n = 40	Enzalutamide n = 36	P value
PSA50 response, n (%)	20 (50)	14 (39)	0.75
Median TTF, months	24.2	21.9	0.05
Median OS, months	31.6	28.2	0.19
Adverse events (AEs),an (%)	15 (37.5)	11 (31)	0.79
Falls	3 (7)	6 (17)
Cardiac failure	2 (5)	1 (3)
Fatigue	2 (5)	1 (3)
Liver function derangement	1 (3)	0
Fracture other	1 (3)	0
Treatment cessation due to AE, n (%)	6 (15)	3 (8)	0.48

ePAD, electronic Prostate Cancer Australian Database; OS, overall survival; PSA50, ≥50% reduction in prostate-specific antigen from baseline; TTF, time to treatment failure.

^aAdverse events recorded in ePAD are limited to those leading to hospitalisation, treatment cessation, or dose modification.

Figure 3.

Time to treatment failure and overall survival by treatment group. CI, confidence interval.

On univariate analysis (Supplementary Table S2, available at https://doi.org/10.1016/j.esmorw.2025.100677), higher baseline GDS score was associated with inferior TTF (HR 1.3, 95% CI 1.05-1.6, P = 0.014), as was higher baseline FRQ score (HR 1.12, 95% CI 1.09-1.22, P = 0.012) and higher 12-week GDS score (HR 1.30, 95% CI 1.05-1.62, P = 0.018) (Supplementary Table S3, available at https://doi.org/10.1016/j.esmorw.2025.100677). Older age was also associated with inferior TTF on univariate analysis (HR 1.12, 95% CI 1.05-1.21, P = 0.001). Hypertension was also associated with TTF in univariate (HR 0.46, 95% CI 0.22-0.94, P = 0.035) but not multivariate modelling (HR 0.53, 95% CI 0.20-1.42, P = 0.21).

Baseline GDS was associated with inferior OS on univariate analysis (HR 1.4, 95% CI 1.12-1.76, P = 0.004) and this was maintained on multivariate analysis (HR 1.37, 95% CI 1.02-1.85, P = 0.037) (Supplementary Table S2, available at https://doi.org/10.1016/j.esmorw.2025.100677). Increasing age was associated with poorer OS on univariate but not multivariate analysis (univariate HR 1.08, 95% CI 1.01-1.63, P = 0.028; multivariate HR 1.12, 95% CI 0.99-1.34, P = 0.07). Higher Charlson score was associated with inferior OS only on univariate analysis (HR 1.79, 95% CI 1.06-3.02, P = 0.029). Neither the randomisation to ARPI nor prior docetaxel was significantly associated with TTF or OS.

Discussion

REAL-Pro is the first RRCT in advanced prostate cancer to our knowledge, and the first randomised study comparing outcomes in older patients receiving ARPIs for mCRPC. While our study had small numbers and was underpowered to detect significant differences, we demonstrated the potential applicability of abbreviated assessment tools to guide clinical practice. Patients aged ≥75 years receiving enzalutamide had numerically higher FRQ scores after 12 weeks of ARPI treatment and a trend of higher mean score changes from baseline for BOMC, FRQ, and GDS compared with those receiving abiraterone. On univariate analysis, baseline GDS and FRQ scores were associated with TTF and OS, while higher 12-week GDS scores were associated with TTF. The 12-week FRQ scores also correlated with the incidence of falls. Importantly, we also highlighted several challenges relating to the conduct of RRCTs, common to standard RCTs, which ultimately led to early closure of this study. Our experience and insights will enable improvement of future RRCT trial design and implementation.

A recent presentation of the COGCaP, comparing patient-reported outcomes prospectively in a non-randomised cohort of all ages receiving abiraterone or enzalutamide, also demonstrated slow accrual. They enrolled 74 of a planned 100 patients, with no differences in cognitive assessments detected at the 3-month timepoint.²⁶ Similarly, the non-randomised ACE study demonstrated comparable composite cognitive outcomes in patients of all ages treated with abiraterone compared with enzalutamide for mCRPC. However, those receiving enzalutamide had greater fatigue, depression, and perceived cognitive deficits.²⁷ This is consistent with our randomised study findings, noting that our study population was likely to be more susceptible to AEs, given a greater incidence of baseline comorbidities and functional impairment in older patients.

There are extensive data supporting the use of CGAs, demonstrating correlation with treatment toxicity and the risk of hospital admissions.^13,17 International guidelines recommend these assessments are carried out in all patients aged >65 years, but these are time and resource intensive, hence are not feasible in many real-world oncology settings.²⁸ Studies suggest that CGAs are used in <50% of older cancer patients, with time constraints being a common barrier.28, 29, 30 Screening assessments using validated tools, such as those used in our study, may therefore be more practical and help to prioritise those who require further comprehensive testing or monitoring. This approach has also been suggested in the most recent American Society of Clinical Oncology (ASCO) guidelines.¹⁷

Our study demonstrated a 74% completion rate for 12-week questionnaires and detected changes at this early timepoint. These completion rates are lower than those anticipated in our initial sample size calculation, where a 10% attrition rate was proposed. These rates are however consistent with those reported in standard RCTs, which vary substantially and generally decrease over time. One review of the European Organisation for Research and Treatment of Cancer trials demonstrated completion rates between 23% and 92%.³¹ Our completion rate demonstrates the feasibility of integrating these tools into RRCTs, and to clinical registries to allow real-time feedback of results to clinicians, supporting early intervention and improved patient-centred care.

In our study there was no significant difference in treatment efficacy between treatment groups, including PSA response rate, TTF, and OS. Two randomised trials have demonstrated greater PSA response rates with enzalutamide compared with abiraterone in mCRPC, but similar time to progression⁵ and OS.⁶ Both studies included patients of all ages. Grade 3-4 AEs were more common with abiraterone, although enzalutamide was associated with more fatigue. Several retrospective analyses have suggested superior OS with enzalutamide compared with abiraterone, likely influenced by selection bias, increased cardiovascular risk and complications in the real-world setting.^7,32 However, a retrospective analysis of older patients demonstrated more dose reductions leading to earlier treatment discontinuation with enzalutamide compared with abiraterone but similar survival outcomes.³³ Furthermore, a recent meta-analysis of individual patient data from pivotal HSPC trials has suggested a smaller magnitude of benefit for ARPIs in patients aged ≥75 years, including no benefit with abiraterone in the oldest subgroup.³⁴

Median OS in our study was lower than seen in the pivotal PREVAIL and COU-AA-302 studies, which included 35% and 32% of patients aged ≥75 years, respectively.^35,36 Given that previous analyses demonstrated that 43% of patients within ePAD were ≥75 years of age at the time of CRPC diagnosis, the greater number of comorbidities present in this real-world population is likely to have influenced survival. The rate of AEs was higher in the older patient subgroup within PREVAIL and COU-AA-302. Older patients receiving enzalutamide had higher grade 3 AEs (48.9% versus 39.5%), more falls (13.8% versus 5.6%), and more fractures (15.8% versus 9.9%). In our study, the rate of falls was 17% but the overall AE rates were lower than those in the pivotal trials, likely reflecting differences in registry versus clinical trial definitions.

Optimising study designs for RCTs, including finding a question with clinical equipoise, is vital for successful implementation. The slow recruitment rate for REAL-Pro likely reflects a lack of equipoise, given the real-world incidence of comorbidities and polypharmacy leading to patients or clinicians favouring one ARPI over the other. Subsequent retrospective analyses of ePAD data have highlighted that potential comorbidity interactions with abiraterone were identified in 72% of patients within a subset of ePAD patients and 14% with enzalutamide, while clinically significant drug interactions with abiraterone were present in 44% and with enzalutamide in 67% of patients.²⁵

Previous studies had suggested that enzalutamide was associated with greater neurocognitive deficits in all patients and inferior health-related quality of life in those aged ≥75 years,^10,11 despite similar efficacy. Our data had also suggested a higher rate of enzalutamide use compared with abiraterone in our registry population.³⁷ We therefore aimed to conduct an RRCT in the older, typically underrepresented population,³ to answer a clinically meaningful question. Our results were consistent with these previous RCTs and retrospective studies.

The small sample size in REAL-Pro was a significant unforeseen limitation, resulting from the changing treatment landscape and reduced clinical equipoise following emerging data during the recruitment period. This limits the statistical power and therefore much of our data demonstrate interesting trends but were not statistically significant. Despite this, the suggestion of differing toxicity profiles and the potential integration of geriatric screening tests that can practically be applied into clinical practice warrant further study. Although recruitment was limited by the change in treatment landscape with ARPIs being more commonly used in the HSPC setting, our results are still relevant as the differences in toxicity profiles and impact on patient-reported outcomes will have increasing importance in this earlier setting, as patients are more likely to remain on treatment for longer periods. Given ARPIs are now accessible as standard-of-care treatments in the HSPC setting, including darolutamide, which is associated with fewer neurocognitive side-effects,³⁸ we plan to analyse the impact of comorbidities and drug interactions as well as patient-reported assessments in the real-world ePAD HSPC registry cohort.

Another limitation is that the patient-reported outcome assessment tools may not be sensitive enough to detect subtle but important differences. The tools used in our study had been validated in oncology settings and were chosen as part of this pragmatic study design, with potential feasibility to be integrated into real-world clinical care. Although more time consuming, CGAs are likely to cover more domains for assessment of health-related quality of life. For example, the BOMC does not test executive function, which may be an early subtle sign of cognitive impairment affecting quality of life. Furthermore, although prior studies have suggested that cognitive changes can be detected early, cumulative effects may further impact cognition and quality of life beyond the timeframe of our study. However, the trends detected and association of certain scores with TTF and OS in our study, despite these limitations, highlight the potential value of these tools in screening for vulnerable patients in the real-world setting.

While the need for pragmatic trial designs and rationalising of the frequency of assessments is a limitation, it can also increase trial participation, particularly recruitment from smaller and geographically dispersed sites that often have fewer resources, leading to a trial population that is more representative of the real-world population. A qualitative study³⁹ has identified barriers to RRCTs, however, including time constraints in explaining trial concepts, particularly in the setting of other competing interventional trials and the fact that RRCTs are often considered ‘less exciting’. The study participants suggested that greater education as well as incentives may promote further recruitment. However, there is a limited understanding about the perspectives of those who declined to participate in RRCTs, including patients and clinicians. Therefore, a subsequent qualitative sub-study is planned to explore clinician and patient perspectives of RRCTs, including those who previously declined participation in REAL-Pro, to enable greater understanding and develop strategies to engage investigators and improve recruitment in future trials.

Conclusion

Our results suggest differences in ARPI toxicity profiles within a real-world cohort of older men with mCRPC. The association with lower scores with long-term outcomes highlights the importance of personalised assessments of cognition, depression, and falls risk in older men before and during ARPI treatment.

While registry-based trials are pragmatic and offer several advantages in increasing inclusivity and generalisability in this underrepresented population, our experience in conducting REAL-Pro also highlights the potential challenges regarding implementation and recruitment. The slow accrual in REAL-Pro was likely related to lack of equipoise in treatment selection. A planned qualitative sub-study will increase understanding of clinician and patient perspectives to identify barriers and develop strategies to improve future RRCT design and implementation.